Abstract

A comparison of cause specific standarised mortality ratios (SMRs) and proportionate mortality ratios (PMRs) or proportionate cancer mortality ratios (PCMRs) was made based on the mortality experience of a cohort of 34 156 members of a heavy equipment operators union. Two types of PMRs or PCMRs were used in the comparison: those based on all deaths and those based on deaths known to the union only. The comparison indicated that, for the entire cohort, both types of PMRs were poor indicators for cancer risk and produced a large number of false positives. On the other hand, PCMRs appeared to be better than PMRs for assessing the direction of site specific cancer risk, but they tended to overstate the magnitude of risk. Analysis by duration of union membership or latency indicated that PMRs or PCMRs based on deaths known to the union tended to overestimate the risk of lung cancer by disproportionately larger amounts in groups with shorter time than in groups with longer time. This differential bias had the net effect of reducing the gradient of any trend or eliminating the trend entirely. In conclusion, PMR or PCMR, based on reasonably sufficient death ascertainment, has a certain usefulness in generating hypotheses, but they are not useful or reliable in measuring the magnitude of risk or in detecting trends in dose response analysis. No conclusion should be drawn from either PMR or PCMR.